Low-Emission Combustion
The attainment of very low pollutant emissions, in particular oxides of nitrogen
( NOx), from gas turbines is not only of considerable environmental concern but has
also become an area of increasing competitiveness between the different engine
manufacturers. For stationary engines, the attainment of ultralow NO~ has become
the foremost marketing issue. This paper is devoted primarily to current and emerging
technologies in the development of ultralow emissions combustors for application to
aircraft and stationary engines. Short descriptions of the basic design features of
conventional gas turbine combustors and the methods of fuel injection now in wide-
spread use are followed by a review of fuel spray characteristics and recent develop-
ments in the measurement and modeling of these characteristics. The main gas-
turbine-generated pollutants and their mechanisms of formation are described, along
A. H. Lefebvre with related environmental risks and various issues concerning emissions regulations
Reilly Professor Emeritus of and recently enacted legislation for limiting the pollutant levels emitted by both
Combustion Engineering, aircraft and stationary engines. The impacts of these emissions regulations on com-
Thermal Science and Propulsion Center, bustor and engine design are discussed first in relation to conventional combustors
Purdue UniversLty, and then in the context of variable-geometry and staged combustors. Both these
West Lafayette, IN 47907 concepts are founded on emissions reduction by control of flame temperature. Basic
approaches to the design of "dry" low-NOx and ultralow-NOx combustors are re-
viewed. At the present time lean, premix, prevaporize combustion appears to be
the only technology available for achieving ultralow NOx emissions from practical
combustors. This concept is discussed in some detail, along with its inherent problems
of autoignition, flashback, and acoustic resonance. Attention is also given to alterna-
tive methods of achieving ultralow NOx emissions, notably the rich-burn, quick-
quench, lean-burn, and catalytic combustors. These concepts are now being actively
developed, despite the formidable problems they present in terms of mixing and
durability. The final section reviews the various correlations now being used to
predict the exhaust gas concentrations of the main gaseous pollutant emissions from
gas turbine engines. Comprehensive numerical methods have not yet completely
displaced these semi-empirical correlations but are nevertheless providing useful
insight into the interactions of swirling and recirculating flows with fuel sprays, as
well as guidance to the combustion engineer during the design and development
stages. Throughout the paper emphasis is placed on the important and sometimes
pivotal role played by the fuel preparation process in the reduction of pollutant
emissions from gas turbines.
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 617
Copyright © 1995 by ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
• Basic approaches to the design of " d r y " low NOx and ,,PRIMAR.y.=L__INTERMEDIATE_~.L__ DILUTION____,=
ultralow NOx combustors, with emphasis on the need to FUEL ZONE
T ZONEZ O N E
feed the combustion zone with completely homogeneous
fuel-air mixtures, with the attendant problems of autoig- N
I JECTOR~ I (
nition and flashback.
• Alternative methods for achieving ultralow NOx emis-
sions, notably rich-burn, quick-quench, lean-burn, and
catalytic combustors, both of which combine high poten-
tial for very low levels of nitric oxides with formidable
mixing and mechanical design problems.
• Correlation and modeling of nitric oxides and carbon
monoxide emissions.
¢3 1.2
0.8 / ~ 0 I.tm,q - 2
~ SllcOndl~NoIzle
0.4
Fig. 2 Dual-orifice, pressure-swirl atomizer (courtesy of the Parker
Hannifin Corporation)
00 I I ~ I ~l I
50 100 150 200 250
DROPDIAMETER,IJ.rn
Gas Injection. For almost all stationary gas turbines the
fuel employed is natural gas. Some engines are fitted with dual- Fig.4 Influenceof dropsizedistributionparameteronfrequencydistri-
fuel nozzles, which allow both liquid and gaseous fuels to be butioncurve
used, but usually the liquid fuel is for backup purposes only.
The methods used to inject gaseous fuels tend to be simple as
there are no requirements for atomization and fuel evaporation. rived from basic principles. In consequence, the majority of
They include injecting the gas through plain orifices, slots, investigations into the drop sizes produced in atomization have
swirlers, and venturi nozzles. resulted in empirical equations for mean drop size. Lefebvre
The main problem with gas injection is that of achieving (1989) has reviewed the available equations for the mean drop
the optimal level of mixing between air, fuel, and combustion sizes produced by the types of atomizers employed in gas tur-
products in the combustion zone. Too high a mixing rate pro- bine combustion. They show that mean drop sizes are largely
duces narrow burning limits. On the other hand, if the rate of dependent on atomizer size, design features, and operating con-
mixing is too low, the system may be prone to low-frequency, ditions, and also on the physical properties of the liquids em-
combustion-induced pressure oscillations. On engines designed ployed and the density of the surrounding gaseous medium.
to operate on both gaseous and liquid fuels, it is important that Most of the equations for mean drop size published before
the gas flow pattern be matched to that of the liquid fuel; other- around 1970 should be regarded as suspect due to deficiencies
wise some variation in the temperature distribution of the com- in the methods available for drop size measurements. Even
bustor outlet gases would occur during changeover from one expressions based on accurate experimental data should only
fuel to the other. be used within the ranges of air properties, liquid properties,
and atomizer operating conditions employed in their derivation.
Fuel Spray Characteristics Extrapolation to other conditions is fraught with risk because
changes in any of these variables could produce a change in
The spray properties of most relevance to the formation of the mode of atomization, which could have a significant effect
pollutant emissions are mean drop size, drop size distribution, on the manner and extent to which variations in the relevant
cone angle, and patternation. Other properties of special impor- flow parameters affect the drop size distribution in the spray.
tance for the successful modeling of spray characteristics in-
clude droplet and gas velocities, droplet trajectories, and mass Drop Size Distribution. Owing to the random and chaotic
flux distributions. nature of the atomization process, the threads and ligaments
formed by the various mechanisms of jet and sheet disintegra-
Mean Drop Size. Various definitions of mean drop size tion vary widely in diameter, and their subsequent breakup
are available, of which the most widely used is the Sauter mean yields a correspondingly wide range of drop sizes. Most practi-
diameter (SMD), which represents the volume/surface ratio of cal atomizers produce droplets in the size range from a few
the liquid in the spray. Unfortunately, the physical processes microns up to several hundred microns.
involved in atomization are not yet sufficiently well understood As no complete theory has yet been developed to describe
for mean diameters to be expressed in terms of equations de- the hydrodynamic and aerodynamic processes involved when
jet and sheet disintegration occurs under normal atomizing con-
ditions, a number of functions have been proposed, based on
either probability or purely empirical considerations, that allow
the mathematical representation of measured drop size distribu-
tions. The simplest expression for drop size distribution is that
of Rosin and Rammler (1933). It may be expressed in the form
~ F shr°ud
1 - Q = exp - ( D / X ) q (1)
uterAit Swider
where Q is the fraction of the total volume contained in drops
Inner Ai uel Swider of diameter less than D, and X and q are constants. Thus, the
Rosin-Rammler relationship describes the drop size distribu-
tion in terms of the two parameters X and q. The exponent q
provides a measure of the spread of drop sizes. The higher the
value of q, the more uniform is the spray.
Figure 4 shows frequency distribution plots for two sprays,
both having the same SMD of 50 /zm, but in one case the
Fig.3 Prefllming airblast atomizer (courtesy of the Parker Hannifin Cor- distribution parameter q has a value of 2 and in the other case
poration) a value of 3. This figure clearly demonstrates that knowledge
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 619
J i ",,
swirl atomizer design has been toward lower values of It~do in
order to reduce internal losses and thereby improve atomization.
Most current atomizer designs have values of lo/do of 0.5 or
Rp= 9.7mm A
less. Figure 5 clearly demonstrates that reducing It~do can appre-
ciably worsen the spray patternation.
~.1NOZZLE
VENTURI
Spray Diagnostics. In recent years considerable advances RADIALSWlRLER
have been made in the development of laser diagnostic tech- FLARE
niques for measuring the properties of sprays. The methods
used in spray analysis have been reviewed by Chigier (1983),
Chigier and Stewart (1984), and Durst et al. (1981). Laser-
Doppler velocimetry and phase-Doppler anemometry, inter-
ECCENTRIC
faced with signal processing and high-speed data acquisition
systems, are now being used routinely in simultaneous measure- . .,.ARYsw,RL
ments of drop size, drop velocity, and local drop drag coeffi- + +++.!"
..,I=" ®
cients (Bauckhage, 1988; Delarosa et al., 1993; Dittmann et al.,
1992; Naqwi et al., 1991). High-speed pulsed microphotogra- I~ ....... 'R- ,/~1~/~ SECONDARY
SW,RL
phy, cinematography, and holography are also used to study
drop size distributions and spray structure (Santangello and
Sojka, 1994). Much of the tedium normally involved in detailed
tz+ "
studies of drop size distributions in various regions of the spray • NOZZLE
can now be alleviated using automatic image analysis, in which VENTURI
droplet images are enlarged, counted, and sorted by electronic (3)RADIALSWlRLER
FLARE
scanning devices and analyzed by microprocessors. 0.125"
Many efforts are now underway to gain a better understanding
of the basic phenomena occurring within complex practical sys- Fig. 6 Concentric and eccentric airblast atomizers
tems currently in service on advanced turbojet engines. Wang
et al. (1992a, b, 1994, 1995) and McDonell et al. (1994a, b)
used a two-component phase-Doppler interferometry system to downstream. The slight flow asymmetries observed with con-
make detailed measurements in the flows downstream of a centric alignment of the swirlers is attributed to normal hard-
SNECMA/GE CFM-56 airblast atomizer. In addition to time- ware tolerance. The measurements obtained with the eccentric
averaged properties, transient phenomena were also examined. swirler assembly exhibit asymmetries in both mean velocities
These studies have provided great insight into the general struc- and turbulence.
ture of the gas phase as well as details regarding the complex Hebrard et al. (1991, 1993) also used two-component phase-
behavior of drops. Of special interest from a practical viewpoint Doppler interferometry to characterize the two-phase flow field
is that actual engine hardware was used to examine the sensitiv- generated by an engine airblast atomizer. Their measurements
ity of the spray structure to scaling and hardware variations. included mean axial and transverse velocities for each phase as
One study was devoted to an examination of the effects of well as volume flux and drop size distributions. They used
nozzle defects on spray characteristics. Figure 6 presents a sche- surface plots to illustrate the local variation of Sauter mean
matic diagram of the CFM-56 atomizer and swirl cup used in diameter for cross sections of the spray at several different
this investigation. It features a pair of concentric air inlets, distances downstream from the atomizer. Close to the nozzle,
which generate two separate swirling airflows in opposite direc- they observed a dome-shaped distribution of drop sizes. Air
tions. The swirl cup acts as a "hybrid" injector, with drops flowing through the inner and outer swirlers of the airblast
produced from a combination of airblast and pressure atomiza- atomizer produced a nearly uniform drop size distribution, but
tion. This figure illustrates what can happen to the swirl cup recirculating air jets caused some secondary atomization. Some
assembly during the engine operating cycle. The design calls large drops were observed in the core, generated by the inner
for concentric alignment at altitude cruise conditions. However, swirling air. Farther downstream the drop size distribution
due to "throttle pushes" over the years, the assembly can be- broadened and indentations appeared on the spray boundary,
come eccentric by up to around 3 mm (0.125 in.), as indicated caused by the presence of air jets emanating from the holes
in Fig. 6. This occurs because the secondary swirler is attached located near the edge of the atomizer cup.
to the combustor liner, whereas the burner feed arm is mounted Wynne and Jasuja (1991) and Jasuja and Lefebvre (1994)
directly onto the combustor casing. Since the two components used various nonintrusive techniques, including phase-Doppler
are pinned at two different locations to the casing, the thermal interferometry, high-intensity spark photography, high-speed
expansion of the casing causes misalignment. cinephotography, and video imaging to capture the dynamic and
Results showing the effect of misalignment on the flow field unsteady spray characteristics produced in prefilming airblast
downstream of the atomizer are presented in Fig. 7. In this atomization. All of the injectors used in this program were
figure the two circles on each cartoon represent the primary typical of modern gas turbine practice and differed from each
and secondary swirlers. The vector quantities represent the gas other only in regard to various design features, such as swirler
velocities in the r - Z plane, while the contours are the turbulent flow area and vane angle, m special feature of this work is that
kinetic energies based on the measured rms velocity compo- it covers a very wide range of ambient air pressures. The results
nents. The highest velocity variations occur in the shear layer obtained are generally consistent with the findings reported
between the recirculating flow and the surrounding flow moving above in showing that the sprays produced by actual engine
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 621
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 623
Emissions Regulations This equation demonstrates that improving (i.e., reducing) the
specific fuel consumption results directly in a reduction of the
The International Civil Aviation Organization (ICAO) has pollution levels. Because the CO and UHC levels of modem
promulgated regulations for civil subsonic turbojet/turbofan en- engines have been greatly reduced at all low power conditions,
gines with rated thrust levels above 26.7 kN (6,000 pounds) and only NOx is emitted in appreciable amounts at altitude
for a defined landing-takeoff cycle (LTO) which is based on cruise, in practice the emissions generated by aircraft engines
consist primarily of NOx. A typical example of the emissions
Table 2 ICAO gaseous emissions standards mass distribution associated with the flight of a modem subsonic
aircraft has been provided by Bahr (1992) and is shown in
Emission, Subsonic turbojet/ Supersonic turbojet/ Table 3. This table represents a flight of 900 km (500 nautical
g/kN turbofan engines* turbofan engines miles) and shows that NOx emissions predominate both in the
vicinity of the airport and during altitude cruise. For a longer
flight, NOx emissions would account for an even larger fraction
o I-K2 19.6 140(0.92)~r,0o of the total emissions mass.
o CO 118.0 4550(~oo)-I.03 The ICAO standard for smoke measurement is expressed in
terms of a Smoke Number (SN), which is related to the engine
o NOx 40+2goo 36+2.421r.oo takeoff thrust (F00) by the expression
*Newly manufactured engines with rated takeoff SN = 83.6(F0o) -°'274 (4)
thrust greater than 26.7 kN.
This expression is shown graphically in Fig. 8. The intention
~oo = engine pressure ratio at takeoff of this standard is to eliminate any visible smoke from the
engine exhaust. As smoke visibility depends on both the smoke
•
50
Ul 40
II1
z 30 el = 83.6 (Fo0)-0.274
W
0
I I co
10
I"i ~ H C
I I
I I I I I I IDLE MAX
0 POWER
0 50 100 150 200 250 300
RATED THRUST (F0o), kN Fig. 10 Emissions characteristics of gas turbine engines
i
. ~ 40
60 NOX
(GAS)
• CO
and smoke are fairly insignificant at low power settings and
attain maximum values at the highest power condition. These
characteristic trends are sketched in Fig. 10.
(LIQUID) Carbon Monoxide. In combustion zones designed to oper-
o> 2o ate fuel-rich, large amounts of CO are formed owing to the lack
• CO
10 (GAS) of sufficient oxygen to complete the reaction to CO2. If, how-
0 ever, the combustion zone mixture strength is stoichiometric or
>50 MW <50 MW US Japan
moderately fuel-lean, significant amounts of CO will be present
EEC DIRECTIVE (19 AUG 92)
owing to the dissociation of CO2. In principle, it should be
Fig. 9 Some emissions regulations for industrial engines possible to reduce this CO to a very low level by the staged
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 625
10-
Influence oflnlet Air Temperature. The effect of an increase
o T A = 473 K in inlet air temperature is to reduce CO, as illustrated in Fig.
8 E1 T A . 573 K 11. This it does by raising the flame temperature, which acceler-
ates the conversion of CO into CO2. At equivalence ratios of
6 stoichiometdc and above, this higher flame temperature pro-
I I I I I I I I
0.6 0.8 1.0 1.2 1.4 motes the formation of CO by dissociation so that, at the highest
equivalence ratios, the CO concentrations are highest for the
EQUIVALENCE RATIO, higher inlet air temperature. These trends are clearly illustrated
in Fig. 11.
Fig. 11 Influence of inlet air temperature on CO emissions
Influence of Pressure. Figure 13 demonstrates the beneficial
effect of an increase in combustion pressure in reducing CO
admission of additional air downstream of the primary zone to emissions. At low equivalence ratios, increase in pressure di-
achieve a gradual reduction in burned gas temperature. In prac- minishes CO by accelerating the rate of conversion of CO into
tice, CO emissions are found to be much higher than predicted CO2. At high equivalence ratios, increase in combustion pres-
from equilibrium calculations and to be highest at low-power sure reduces CO emissions, albeit to a lesser extent, by sup-
conditions, where burning rates and peak temperatures are rela- pressing chemical dissociation.
tively low.
Influence of Ambient Air Temperature. Hung and Agan
The main factors influencing burning rates, and hence also
(1985) have examined the influence of ambient air temperature
CO emissions, are engine and combustor inlet temperatures,
on the CO emissions from a 7 MW industrial engine supplied
combustion pressure, and primary-zone equivalence ratio. (Note
with natural gas fuel. A strong air temperature effect on mea-
that equivalence ratio is the actual fuel/air ratio divided by the
sured CO was observed. CO emissions for an air temperature
stoichiometric value of fuel/air ratio.) All these aspects have
been investigated by many workers, including Bahr (1982,
1987, 1991, 1992), Bayle Labour6 (1991), Correa (1991),
Desaulty ( 1991 ), Koff (1993), Mosier and Pierce (1980), and 40
Sturgess et al. (1993). In their studies on pollutant emissions
from continuous flow combustors, Rink and Lefebvre (1989a) P&W JT9D-70A
used a continuous flow tubular combustor, 150 mm in diameter, 85 IDLE
in conjunction with an array of 36 equally spaced "micro-
scopic" airblast atomizers, to achieve a uniform distribution of
fuel in the mixture entering the combustion zone. This method 30 -
of fuel injection had another useful advantage in that it allowed
the mean drop size in the fuel spray to be varied in a controlled ._
manner while maintaining all other flow conditions constant. ~. 25 -
All measurements of pollutant emissions were carried out at a 9)
distance of 170 mm from the fuel injectors. 5,
Influence of Equivalence Ratio. Some of the results ob- ~ 20-
O
tained by Rink and Lefebvre are presented in Fig. 11, which
shows the variation of CO emissions with equivalence ratio for
two values of inlet air temperature, namely 473 and 573 K. 15 -
Both curves exhibit the same general characteristics. They show
that CO emissions diminish with increase in equivalence ratio,
reaching a minimum value at an equivalence ratio, ~b, of around 10 - O
0.8. Beyond this point any further increase in equivalence ratio
causes CO emissions to rise. These trends are typical of those
observed for other types of combustion systems. The high levels 5 = I I I I
of CO at low equivalence ratios are due to the slow rates of 0.4 0.6 0.8 1.0 1.2 1.4
oxidation associated with low combustion temperatures. In- PRIMARY ZONE EQUIVALENCE RATIO
crease in equivalence ratio raises the flame temperature, which
accelerates the rate of oxidation so that CO emissions decline. Fig. 12 Influence of primary zone equivalence ratio on co emissions
~( 0 0
2C
I I
10 ' I0 I 2~ 50
WEAK SMOKE POINT
LIMIT
Fig. 17 Correlation of smoke number with smoke point for a TF 33
combustor
%
COMBUSTION PRESSURE
ally high quality of the data fit obtained with these and several
Fig. 15 Influence of pressure on flammability limits other aircraft combustors led Chin and Lefebvre to conclude
that smoke point is superior to hydrogen content as a correlating
parameter for soot-related combustion phenomena.
Influence of Fuel Type. Fuel properties can influence smoke
production in two ways: first by inducing the formation of local Oxides of Nitrogen. Some of the nitric oxide (NO) formed
fuel-rich regions, and second, by exerting variable resistance to in combustion can subsequently oxidize to NO2. For this reason
carbon formation. The former is controlled by physical proper- it is customary to lump NO and NO2 together and express results
ties, such as viscosity and volatility, which affect the mean drop in terms of oxides of nitrogen (NOx), rather than NO. At high
size, penetration, and rate of evaporation of the fuel spray, emission levels the predominant compound is nitric oxide,
whereas the latter relate to molecular structure. It is well estab- which is produced mainly by the oxidation of atmospheric nitro-
lished that smoking tendency increases with a reduction in hy- gen in high-temperature regions of the flame. The process is
drogen content and, in fact, hydrogen content is commonly used endothermic and it proceeds at a significant rate only at tempera-
in correlating rig and engine test data on various soot-related tures above around 1850 K. It can be produced by four different
parameters such as smoke emissions, flame radiation, and liner mechanisms, as discussed below.
wall temperature. However, Chin and Lefebvre (1993) have 1 Thermal NO. This is produced by oxidation of atmo-
shown that a better index of sooting tendency is the ASTM spheric nitrogen in the postflame gases. Most of the proposed
smoke point, which is obtained experimentally by burning the reaction schemes for thermal NO utilize the extended Zeldovich
test fuel in a wick lamp and slowly increasing the height of mechanism:
the flame until it begins to smoke. The height of the flame in
millimeters is the smoke point; the higher this is, the lower is N2 + O = NO + N
the tendency of the fuel to soot formation.
N + 02 = NO + O
The correlations shown in Figs. 17 and 18 were obtained
from measurements of smoke number carried out on Pratt and N + OH = NO + H
Whitney TF33 and F100 combustors, respectively. The gener-
NO formation is found to peak on the fuel-lean side of stoichio-
metric. This is a consequence of the competition between fuel
200 and nitrogen for the available oxygen. Although the combustion
temperature is higher slightly on the rich side of stoichiometric,
the available oxygen is then consumed preferentially by the
I00 fuel. For even richer mixtures, a further reduction of NO may
50
z F I00 CRUISE
0d 30 P:5"H2-LI8 MPo
FAR,0.020
co
<t
o
bJ
IC ! I I I I I
I0 20 50
0.5 IO 20 SMOKE POINT
INLET AIR PRESSURE, MPo
Fig. 18 Correlation of smoke number with smoke point for an F 100
Fig. 16 Data illustrating the effect of pressure on exhaust smoke combustor
30 LIQUID FUEL
20 45C
~. lO
3OC
15C
I I I I
06 0.8 ID 1.2
EQUIVALENCE RATIO
11/050 I
1600 I
1700 I
1800 I
1900 I
2000
Fig. 20 Influence of inlet air temperature on nitric oxide emissions
FLAME TEMPERATURE, K
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 629
N2 + C H = H C N + N
The balance of the prompt NOx mechanism involves the oxi- Such extrapolation could be carried out with confidence if the
dation of the HCN molecules and N atoms. Under lean-pre- relationship between NOx and pressure were accurately known.
mixed conditions, the HCN oxidizes to NO mainly by a se- Unfortunately, the experimental data obtained on different com-
quence of reactions involving HCN - , CN --* NCO ~ NO. The bustor types are conflicting in this regard. They vary from no
N atom reacts mainly by the second Zeldovich reaction. effect of pressure on NOx to quite significant increases in NOx
Of special interest and importance is the influence of pressure with increase in pressure.
on prompt NO formation because this could be the main contrib- For conventional combustors it is generally found that NOx
utor to the NO emissions produced in lean premixed combustion c~ pn, where n has values ranging from around 0.5 to around
(Correa, 1991 ). There are, however, few data available on this 0.8. Laboratory burners can also exhibit a strong effect of pres-
effect. Fennimore's (1971) pioneering study of prompt NO in sure on NOx. For example, the results of Maughan et al.. (1994)
ethylene-air flames over a range of pressures from 1 to 3 atm from a well-mixed combustor supplied with natural gas fuel
concluded that prompt NO ~ p0.5. Later work by Heberling showed that raising the combustor outlet temperature from 1227
(1976) over a much wider range of pressures from 1 to 18 atm to 1310 K caused n to increase from 0.38 to 0.51. Maughan et
showed that prompt NO was independent of pressure. Altemark al. regard this result as evidence that the lowest NOx levels
and Knauber (1987) also concluded that NOx is independent of result from the nitrous oxide and prompt mechanisms which
pressure for equivalence ratios below 0.6. The practical implica- dominate at low temperatures and which are independent of
tions of these findings are discussed below. pressure, whereas the higher NOx levels associated with higher
combustion temperatures are due primarily to thermal NOx
4 Fuel NO. If fuels contain organically bonded nitrogen,
which exhibits a square root dependence on pressure.
then some of this nitrogen will eventually form the so-called
These results and conclusions are fully consistent with those
"fuel NO." The percentage of nitrogen undergoing this change
obtained by Correa and co-workers (Leonard and Corea, 1990;
increases only slowly with increasing flame temperature. Light
Correa, 1991 ). These workers studied turbulent premixed meth-
distillate fuels contain small amounts of organic nitrogen, less
ane-air flames using an uncooled perforated plate burner op-
than 0.06 percent, but the heavy distillates may contain as much
erating at pressures from 1 to 10 atm, inlet air temperatures
as 1.8 percent. Thus, depending on the degree of nitrogen con-
from 300 to 615 K, and equivalence ratios from 0.5 to 0.9.
version, the fuel NO can represent a considerable proportion of
Their modeling featured a stirred reactor for flame stabilization
the total NO (Merryman and Levy, 1975).
followed by a plug flow reactor and a kinetic scheme which
Nicol et al. (1992) have examined analytically the relative
included thermal and prompt NO. The results confirmed that
contributions of the various mechanisms discussed above to the
the low temperatures of lean flames preclude significant forma-
total NOn emissions produced by a lean-premixed combustor
tion of NO by the thermal mechanism. At temperatures below
burning methane fuel, for which, of course, the fuel NO is
1800 K the prompt mechanism appears to be dominant. The
zero. The results of their study showed that at relatively high
implication of these results to the effect of pressure on NO2
temperatures of around 1900 K, and equivalence ratios of
formation is well illustrated in Fig. 22, which contains some of
around 0.8, the contributions are about 60 percent thermal, 10
the experimental data of Correa et al. and which highlights their
percent nitrous oxide, and 30 percent prompt. With reductions
conclusions in regard to the influence of flame temperature on
in temperature and equivalence ratio, the contributions made
the pressure dependence of NO~ formation. In this and in other
by nitrous oxide and prompt NO increase significantly until, at
figures contained in this paper, the values of NO and NOx quoted
a temperature of 1500 K and an equivalence ratio of around
have been corrected to the values that would have been obtained
•0.6, the relative contributions to the total NOx emissions become
if the combustion air had contained 15 percent oxygen. The
5 percent thermal, 30 percent nitrous oxide, and 65 percent
following expression is used to make this correction:
prompt. These results clearly have great importance to the de-
sign of ultralow NOx lean-premixed combustors. (NOx)15%O2 = (NOx) × 5.9/(20.9 - 02 .......) (6)
Influence of Pressure. The influence of pressure on NOx where NOx concentrations are expressed in ppm by volume on
formation is of special importance due to the continual trend a dry basis and 02 content is expressed as a percentage by
toward engines of higher pressure ratio to meet the need for volume on a dry basis.
lower fuel consumption. Combustor testing at high pressures is The purpose of this parameter is partly to remove ambiguity
extremely expensive and it would therefore be highly conve- when comparing different sets of experimental data, but mainly
nient to carry out combustor development at lower levels of to indicate that combustors burning less fuel are expected to
pressure and then extrapolate the results obtained to high levels produce less NOx. Figure 22 shows that NOx is independent
of pressure where NOx emissions attain their highest values. of pressure in the leanest premixed flames. Increase in flame
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 631
E 2.C
o~
Z
100
80 f o_
U)
60
bJ
§ ,o
40 / / / ,SMD, ~ m
u) 0.8
I ip
I dl
/ // IZ1II0
20 0.6
Q5 >.,
I I I I I I
I I | I I I I I I
O8 Q9 I0 I.I 1.2 13
10 0.6 0.8 1.0 1.2 1.4 EQUIVALENCE RATIO
EQUIVALENCE RATIO, Fig. 26 Influence of mean drop size on smoke emissions
•
Royce engines is not due to prevaporization of the fuel but DUAL
rather to the premixing of fuel and air which occurs within the ¢n20 ORIFICE
vaporizer tubes. Z
Alleviating soot formation and smoke by fuel-air mixing is
only fully effective if sufficient air is used. This is well illus- w 15
trated in Fig. 27 from Sturgess et al. (1993), which shows how
smoke was drastically reduced in a P&W JT9D-70 combustor ¢n I0
when operating at takeoff conditions by the addition of more
air through the fuel injector and air swirler. The injection of air
through these components is particularly effective in reducing
smoke because it all flows into the primary combustion zone, ' 9--A'RBLAST,
which is usually deficient in oxygen and therefore has a high 0
140 160 180
tendency toward soot formation.
OVERALL AIR/FUEL RATIO
The advantage of airblast atomizers over dual-orifice pressure
atomizers in regard to smoke emissions is well established. The Fig. 28 Influence of fuel injector type on exhaust smoke
i =
0
have reported NOx reductions of around 60 percent for a steam/
fuel mass ratio of unity on two GE industrial engines burning -10~
natural gas.
Although both water and steam injection are very effective
in reducing NOx emissions, and have been used on stationary - 5
engines that operate at near-constant load conditions since the
early 1970s, they do have a number of drawbacks. White et al.
(1982) have reported an increase in capital cost of 10 to 15 21000
U.S. dollars per kW and an increase in fuel consumption of 2
to 3 percent. This additional fuel is needed to heat the water to PRIMARY TEMPERATURE, K
combustion temperature, although power output is enhanced
due to the additional mass flow through the turbine. The water 1000 K ~ TEMPERATURE RANGE OF : 2500 K
CONVENTIONAL COMBUSTORS
must be of high purity to prevent deposits and corrosion in the
hot sections downstream of the combustor. The treatment of Fig. 31 Influence of primary-zone temperature on CO and NOx emis-
this water is expensive and requires a separate plant based on sions
reverse osmosis and de-ionization. Users' experience with water
injection has shown a significant increase in inspection and
hardware maintenance. There are, therefore, practical limits to weighs 111 tons. Despite these drawbacks, the method is quite
the amount of water or steam that can be injected into the widely used.
combustor. The deterioration in combustion performance aris-
ing from water-steam injection is manifested as increases in the Exhaust Gas Recirculation. It has been known for many
levels of CO and UHC emissions and by increases in combustor years that recirculation of cooled exhaust products into the inlet
pressure oscillations. These oscillations can become amplified of a gas turbine combustor would reduce the emissions of NOx.
by coupling with the combustion process, and cause deteriora- The practical feasibility of this method of NOx reduction has
tion of combustor hardware. been investigated by Wilkes and Gerhold (1980) who found
The drawbacks associated with water and steam injection that significant reductions (50 percent) could be achieved with
have encouraged the development of the so-called "dry low recirculation rates of 20 percent or less at baseload conditions.
NOx" combustors, i.e., combustors that can meet the emission The major thermal effect stems from the reduced concentration
goals without having to resort to diluent injection. of oxygen in the inlet air, but there is also a secondary effect
Selective Catalytic Reduction ( S C R ) . This is a method due to the higher heat capacity of this air with increased H20
for converting NOx in a gas turbine exhaust stream into molecu- and CO2 content.
lar nitrogen and water vapor by injecting ammonia into the The main advantage of the method is that little or no combus-
stream in the presence of a catalyst. Exhaust gases first pass tor development is required and standard production combustors
through an oxidation catalyst and are then mixed with ammonia can be used. Its main drawback lies in the need for an intercooler
before entering the SCR catalyst. The oxidation catalyst re- between the exhaust and inlet. This virtually rules it out for
moves the CO and UHC emissions by oxidizing them to CO2 simple gas turbines, but application to combined cycle plants
and H20. To reduce NOr emissions, ammonia is injected in a offers more promise due to the substantially lower exhaust gas
manner designed to achieve intimate mixing with the exhaust temperatures. Another drawback is that only very clean fuels
stream. After mixing, the exhaust gases pass over a base metal can be used to avoid problems of fouling and contamination.
catalyst which results in the selective reduction of NOx to form
N2 and H20. The principal reactions are
Pollutant Reduction by Control of Flame Tempera-
6NO + 4NH3 ~ 5N2 + 6H20 ture
6NO2 + 8NH3 ~ 7N2 + 12H20 From the foregoing discussion it is abundantly clear that the
main parameter governing pollutant emissions is the tempera-
SCR requires that the temperature of the exhaust stream be ture of the combustion zone. With conventional combustors this
within a fairly narrow range from 560 to 670 K, and so is can range from 1000 K at low power operation to 2500 K at
restricted to systems in which the exhaust gas flows into a heat high power operation, as indicated in Fig. 31. This figure also
recovery device, usually a steam generator (Davis and Washam, shows that too much CO is formed at temperatures below
1989). A major problem with this method is the requirement around 1670 K while excessive amounts of NOx are produced
for a control system that feeds the requisite amount of ammonia, at temperatures higher than around 1900 K. Only in the fairly
and the need for a continuous monitoring system that can give narrow band of temperatures between 1670 and 1900 K are the
the feedback to the ammonia supply mechanism under differing levels of CO and NOx below 25 and 15 ppm, respectively.
load conditions. Another problem is the size and weight of the The underlying principle of the various approaches toward low-
equipment. According to Davis and Washam, for an 83 MW emission combustors described below is that of maintaining the
MS7000 gas turbine an SCR designed to remove 90 percent of combustion zone (or zones) within a fairly narrow band of
the NOx from the exhaust stream has a volume of 175 m 3 and temperatures over the entire power range of the engine.
Journal of Engineering for Gas Turbines and P o w e r OCTOBER 1995, Vol. 117 / 635
. . . .
I I
I I
i
Fig. 33 Principle of staged combustion
Journal of Engineering for Gas Turbines and P o w e r OCTOBER 1995, Vol. 117 / 637
JPllot
O"
C) ~ j t Orifices
C~
Q
I t/ I,/'%.1
ivi Ill
= ~ , Combu=o,
Liner Fer
t
Methanol / I
Supply Swirler ,-.~, '~
/ I J Swirler
Air L-r-----I--I Centerbody
Row Main o] I oine
Fuel I " -
Fig. 38 Solar Iow-NOx fuel injector for liquid fuels
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 639
MIXING
INNER _ ~ DUCT
SWIRLER CENTERBODY
FUEL (a)
IN
. INJECTION HOLES
OUTER
MIXING
INNER , ~ DUCT CENTERBODY
SWIRLER
Fig. 41 General Electric LM6000 dry Iow-NOx combustor FUEL (b)
IN
Journal of Engineering for Gas Turbines and P o w e r O C T O B E R 1995, Vol. 117 / 641
1.2
~m/UD°' S)(/.Lm]
rises, fuel vapor is formed at the drop surface and part of the
heat transferred to the drop is used to furnish the heat of vapor-
~ 5000
ization of the fuel. Eventually, a stage is reached where all of 1.0
the heat transferred to the drop is used as heat of vaporization
and the drop stabilizes at its "wet-bulb" or "steady-state" 0.8
temperature. Thus the total drop evaporation time can be subdi-
100
vided into two main components, one for the heat-up period 0.6 Too= 2000 K o
and another for the steady-state phase.
During the steady-state phase, evaporation rates are relatively "'
high, but during the initial heat-up period much of the heat 1.0
transferred to the drop is absorbed in heating it up, so the amount
of heat available for fuel vaporization is correspondingly less. 0.8
This lower rate of vaporization, when considered in conj unction =
with the significant proportion of the total drop lifetime occu- N"
pied by the heat-up period, especially at high pressures, means ~ 0.6
that overall evaporation rates can be appreciably lower than the ~.
experimental values quoted in the literature, most of which were '~ 0.4
measured during steady-state evaporation at normal atmo- 200
spheric pressure, i Too = 12OO K =o~
In most continuous flow combustors, the fuel is sprayed into 0.2
air or gas flowing at high velocity. Where relative motion exists
between the droplets and the surrounding gas, the rate of evapo- 0.24
ration is enhanced. The effect of convection on evaporation
rates can be accommodated by applying a correction factor to
the evaporation rate calculated for quiescent conditions. Where 0.213
heat transfer rates are controlling, the correction factor is
0.1~
1 + 0.22 Re~ 5
0.I~
where ReD, the drop Reynolds number, is typically around 5.
From a practical viewpoint it would be very convenient if
the effect of the heat-up period could be combined with that of O.Oe
forced convection in a manner that would allow an "effective"
value of evaporation constant to be assigned to any given fuel 0
0.04 I ! I I I I I
at any stipulated conditions of ambient pressure, temperature, 420 440 460 480 500 520 540 56O
velocity, and drop size. To accomplish this, Chin and Lefebvre Tbn, K
(1982) defined an effective evaporation constant as
Fig. 47 Variation of effective evaporation constant with normal boiling
point for a pressure of 100 kPa
horf = D2/te (9)
1
S l I I I I
(1993). These workers examined the effects of incomplete
fuel-air mixing on the lean blowout limits and emissions char-
1650 1700 1750 1800 1850 1900 1950 acteristics of an LPP coaxial mixing tube combustor. A fluores-
FLAME TEMPERATURE, K cence technique was used to determine the degree of fuel vapor-
ization and mixing at the combustor inlet. Contrary to what was
Fig. 48 Effects of nonuniform fuel-air mixing on NO. formation expected, it was found that an improvement in fuel-air mixing
resulted in comparable NOx emissions for the same conditions
of inlet temperature and equivalence ratio. Although the better
into Eq. (10) should, of course, be that of the largest drop in mixed device demonstrated lower NOx emissions, this was at-
the spray. tributed to its ability to operate at lower equivalence ratios.
In a number of recent publications, Chin (1994, 1995a, b) The formidable problems involved in trying to attain com-
has developed and extended the h~ff concept to include several plete homogeneity in the fuel-air mixture entering the combus-
phenomena of practical importance, including the secondary tion zone have prompted studies into the use of mechanical
breakup of droplets as well as real gases, near-critical tempera- mixers as a means of achieving the desired degree of fuel-air
ture, and multicomponent fuel effects. Recent advances in the mixing. Static mixers are well known and widely used in pro-
modeling of droplet vaporization have been reviewed by Peng cess engineering for mixing of both gases and liquids, but as
and Aggarwal (1995). Their review also includes the methodol- yet they appear to have evoked little interest for combustion
ogies currently available for representing droplet motion and applications. Poeschl et al. (1994) examined the mixing capabil-
vaporization history in two-phase flow computations. ity of a commercially available static mixer after a series of
tests using pressure, airblast, air-assist, and multipoint injectors
Mixing. The attainment of perfect mixture homogeneity had failed to provide the right degree of homogeneity. All these
prior to combustion is of paramount importance to the success injection devices had either failed to mix fuel vapor with the
of most types of low-emissions combustors. The influence of air flowing near the duct wall or had deposited liquid fuel on
mixture inhomogeneity on NOx formation has been examined the wall. Tests with the static mixer yielded excellent homoge-
by several workers, both theoretically and experimentally. Ap- neity, with a standard deviation of lower than 5 percent for a
pleton and Heywood (1973) employed a burner in which kero- 2 percent pressure loss. This high degree of homogeneity in the
sine fuel was injected using an air-assist atomizer, which al- combustible mixture is very desirable, not only from an emis-
lowed the degree of fuel-air mixing to be varied by changing sions standpoint, but also because it greatly reduces the possibil-
the atomizing air pressure. For low equivalence ratios, they ity of spontaneous ignition. Fuel-lean mixtures tend to have
found that NOx emissions decreased by a factor of ten as a long autoignition delays, but if imperfections in mixing result
result of better fuel-air mixing. In a similar study by Semerjian in local regions in which the equivalence ratio is higher than
et al. (1979) it was found that poorer fuel-air mixing resulted the average value, the ignition delay time could be greatly re-
in increased NOx emissions for overall fuel-lean flames. Lyons duced. Thus a high degree of mixture homogeneity is essential,
(1981) used a multipoint fuel injector spraying Jet A fuel to not only for the attainment of low NOx emissions, but also to
achieve different equivalence ratio profiles across the diameter alleviate the problems of autoignition and flashback. Flashback
of the flametube. The results showed that spatial nonuniformity is unlikely to occur with static mixers if the velocity through
in equivalence ratio resulted in increased NOx emissions for the mixer is kept high ( >20 m/s). Also, by flattening the veloc-
equivalence ratios below 0.7 and decreased NOx emissions for ity profile, the mixer eliminates the boundary layer along which
near-stoichiometric mixtures. flashback is most prone to occur.
Flanagan et al. (1992) used a simple mixing tube fitted with According to Valk (1994) the static mixer system has good
a bluff-body flameholder at its exit. By changing the location potential for engines of modest compression ratio, but more
of the natural gas fuel injector along the length of the tube, the work is needed to reduce length, residence time, and pressure
degree of fuel-air mixing in the mixture approaching the stabi- loss, while maintaining good mixing performance.
lizer could be varied. When the system was operating at an
equivalence ratio of 0.66, a nearly fivefold increase in NO+ Spontaneous Ignition. Spontaneous ignition, or autoigni-
emissions was recorded when going from well-mixed to incom- tion, is a process whereby a combustible mixture undergoes
pletely mixed conditions. chemical reaction, leading to the rapid evolution of heat in the
Fric (1992) used an experimental apparatus very similar to absence of any concentrated source of ignition such as a flame
that employed by Flanagan et al. to examine the NOx emissions or spark. In the LPP combustor and other types of low-emission
produced when burning natural gas fuels at normal atmospheric combustors in which fuel and air are premixed prior to combus-
pressure. He found that temporal fluctuations in equivalence tion, spontaneous ignition must be avoided at all costs since it
ratio can also raise NOx emissions, in addition to spatial nonuni- could damage combustor components and produce unacceptably
formities. For example, temporal fluctuations of 10 percent re- high levels of pollutant emissions.
suited in a doubling of NOn emissions. Spontaneous ignition delay may be defined as the time inter-
Leonard and Stegmaier (1994) used a gas-fired GE LM6000 val between the creation of a combustible mixture, say by in-
combustor to examine the effects of premixing on NO+ forma- jecting fuel into a flowing air stream at high temperature, and
tion. The results obtained are given in Fig. 48, which shows the onset of flame. In view of their practical importance, mea-
NO+ as a function of average flame temperature for various surements of spontaneous ignition delay time have been con-
degrees of premixing. Nonuniformities are the result of fluctua- ducted for many fuels over wide ranges of ambient conditions
tions in time as well as variations in space. Figure 48 gives data and in a variety of test vehicles, including constant-volume
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 645
i=i-
Fart L o a d
Fuel Injection
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 647
Journal of Engineering for Gas Turbines and Power OCTOBER 1995, Vol. 117 / 653